X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=kernel%2Ffutex.c;h=438701adce2344faab6551f3c833f3bb36c60b05;hb=cac64d00c256e65776d575e82aaf540632b66178;hp=cf0c8e21d1abaae85c17a816dbd42a8ed3132dd5;hpb=06a9ec291b3aec9c7e36af0a10ad2b556bd7e84f;p=safe%2Fjmp%2Flinux-2.6 diff --git a/kernel/futex.c b/kernel/futex.c index cf0c8e2..438701a 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -16,6 +16,9 @@ * Copyright (C) 2006 Red Hat, Inc., Ingo Molnar * Copyright (C) 2006 Timesys Corp., Thomas Gleixner * + * PRIVATE futexes by Eric Dumazet + * Copyright (C) 2007 Eric Dumazet + * * Thanks to Ben LaHaise for yelling "hashed waitqueues" loudly * enough at me, Linus for the original (flawed) idea, Matthew * Kirkwood for proof-of-concept implementation. @@ -48,37 +51,18 @@ #include #include #include +#include +#include +#include +#include + #include #include "rtmutex_common.h" -#define FUTEX_HASHBITS (CONFIG_BASE_SMALL ? 4 : 8) +int __read_mostly futex_cmpxchg_enabled; -/* - * Futexes are matched on equal values of this key. - * The key type depends on whether it's a shared or private mapping. - * Don't rearrange members without looking at hash_futex(). - * - * offset is aligned to a multiple of sizeof(u32) (== 4) by definition. - * We set bit 0 to indicate if it's an inode-based key. - */ -union futex_key { - struct { - unsigned long pgoff; - struct inode *inode; - int offset; - } shared; - struct { - unsigned long address; - struct mm_struct *mm; - int offset; - } private; - struct { - unsigned long word; - void *ptr; - int offset; - } both; -}; +#define FUTEX_HASHBITS (CONFIG_BASE_SMALL ? 4 : 8) /* * Priority Inheritance state: @@ -106,13 +90,14 @@ struct futex_pi_state { * we can wake only the relevant ones (hashed queues may be shared). * * A futex_q has a woken state, just like tasks have TASK_RUNNING. - * It is considered woken when list_empty(&q->list) || q->lock_ptr == 0. + * It is considered woken when plist_node_empty(&q->list) || q->lock_ptr == 0. * The order of wakup is always to make the first condition true, then - * wake up q->waiters, then make the second condition true. + * wake up q->waiter, then make the second condition true. */ struct futex_q { - struct list_head list; - wait_queue_head_t waiters; + struct plist_node list; + /* There can only be a single waiter */ + wait_queue_head_t waiter; /* Which hash list lock to use: */ spinlock_t *lock_ptr; @@ -120,28 +105,24 @@ struct futex_q { /* Key which the futex is hashed on: */ union futex_key key; - /* For fd, sigio sent using these: */ - int fd; - struct file *filp; - /* Optional priority inheritance state: */ struct futex_pi_state *pi_state; struct task_struct *task; + + /* Bitset for the optional bitmasked wakeup */ + u32 bitset; }; /* * Split the global futex_lock into every hash list lock. */ struct futex_hash_bucket { - spinlock_t lock; - struct list_head chain; + spinlock_t lock; + struct plist_head chain; }; static struct futex_hash_bucket futex_queues[1<index, vma->vm_file->f_dentry->d_inode, - * offset_within_page). For private mappings, it's (uaddr, current->mm). - * We can usually work out the index without swapping in the page. + */ +static void get_futex_key_refs(union futex_key *key) +{ + if (!key->both.ptr) + return; + + switch (key->both.offset & (FUT_OFF_INODE|FUT_OFF_MMSHARED)) { + case FUT_OFF_INODE: + atomic_inc(&key->shared.inode->i_count); + break; + case FUT_OFF_MMSHARED: + atomic_inc(&key->private.mm->mm_count); + break; + } +} + +/* + * Drop a reference to the resource addressed by a key. + * The hash bucket spinlock must not be held. + */ +static void drop_futex_key_refs(union futex_key *key) +{ + if (!key->both.ptr) { + /* If we're here then we tried to put a key we failed to get */ + WARN_ON_ONCE(1); + return; + } + + switch (key->both.offset & (FUT_OFF_INODE|FUT_OFF_MMSHARED)) { + case FUT_OFF_INODE: + iput(key->shared.inode); + break; + case FUT_OFF_MMSHARED: + mmdrop(key->private.mm); + break; + } +} + +/** + * get_futex_key - Get parameters which are the keys for a futex. + * @uaddr: virtual address of the futex + * @shared: NULL for a PROCESS_PRIVATE futex, + * ¤t->mm->mmap_sem for a PROCESS_SHARED futex + * @key: address where result is stored. * - * Returns: 0, or negative error code. + * Returns a negative error code or 0 * The key words are stored in *key on success. * - * Should be called with ¤t->mm->mmap_sem but NOT any spinlocks. + * For shared mappings, it's (page->index, vma->vm_file->f_path.dentry->d_inode, + * offset_within_page). For private mappings, it's (uaddr, current->mm). + * We can usually work out the index without swapping in the page. + * + * fshared is NULL for PROCESS_PRIVATE futexes + * For other futexes, it points to ¤t->mm->mmap_sem and + * caller must have taken the reader lock. but NOT any spinlocks. */ -static int get_futex_key(u32 __user *uaddr, union futex_key *key) +static int get_futex_key(u32 __user *uaddr, int fshared, union futex_key *key) { unsigned long address = (unsigned long)uaddr; struct mm_struct *mm = current->mm; - struct vm_area_struct *vma; struct page *page; int err; @@ -187,131 +215,124 @@ static int get_futex_key(u32 __user *uaddr, union futex_key *key) * The futex address must be "naturally" aligned. */ key->both.offset = address % PAGE_SIZE; - if (unlikely((key->both.offset % sizeof(u32)) != 0)) + if (unlikely((address % sizeof(u32)) != 0)) return -EINVAL; address -= key->both.offset; /* - * The futex is hashed differently depending on whether - * it's in a shared or private mapping. So check vma first. + * PROCESS_PRIVATE futexes are fast. + * As the mm cannot disappear under us and the 'key' only needs + * virtual address, we dont even have to find the underlying vma. + * Note : We do have to check 'uaddr' is a valid user address, + * but access_ok() should be faster than find_vma() */ - vma = find_extend_vma(mm, address); - if (unlikely(!vma)) - return -EFAULT; + if (!fshared) { + if (unlikely(!access_ok(VERIFY_WRITE, uaddr, sizeof(u32)))) + return -EFAULT; + key->private.mm = mm; + key->private.address = address; + get_futex_key_refs(key); + return 0; + } - /* - * Permissions. - */ - if (unlikely((vma->vm_flags & (VM_IO|VM_READ)) != VM_READ)) - return (vma->vm_flags & VM_IO) ? -EPERM : -EACCES; +again: + err = get_user_pages_fast(address, 1, 0, &page); + if (err < 0) + return err; + + lock_page(page); + if (!page->mapping) { + unlock_page(page); + put_page(page); + goto again; + } /* * Private mappings are handled in a simple way. * * NOTE: When userspace waits on a MAP_SHARED mapping, even if * it's a read-only handle, it's expected that futexes attach to - * the object not the particular process. Therefore we use - * VM_MAYSHARE here, not VM_SHARED which is restricted to shared - * mappings of _writable_ handles. + * the object not the particular process. */ - if (likely(!(vma->vm_flags & VM_MAYSHARE))) { + if (PageAnon(page)) { + key->both.offset |= FUT_OFF_MMSHARED; /* ref taken on mm */ key->private.mm = mm; key->private.address = address; - return 0; + } else { + key->both.offset |= FUT_OFF_INODE; /* inode-based key */ + key->shared.inode = page->mapping->host; + key->shared.pgoff = page->index; } - /* - * Linear file mappings are also simple. - */ - key->shared.inode = vma->vm_file->f_dentry->d_inode; - key->both.offset++; /* Bit 0 of offset indicates inode-based key. */ - if (likely(!(vma->vm_flags & VM_NONLINEAR))) { - key->shared.pgoff = (((address - vma->vm_start) >> PAGE_SHIFT) - + vma->vm_pgoff); - return 0; - } + get_futex_key_refs(key); - /* - * We could walk the page table to read the non-linear - * pte, and get the page index without fetching the page - * from swap. But that's a lot of code to duplicate here - * for a rare case, so we simply fetch the page. - */ - err = get_user_pages(current, mm, address, 1, 0, 0, &page, NULL); - if (err >= 0) { - key->shared.pgoff = - page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT); - put_page(page); - return 0; - } - return err; + unlock_page(page); + put_page(page); + return 0; } -/* - * Take a reference to the resource addressed by a key. - * Can be called while holding spinlocks. - * - * NOTE: mmap_sem MUST be held between get_futex_key() and calling this - * function, if it is called at all. mmap_sem keeps key->shared.inode valid. - */ -static inline void get_key_refs(union futex_key *key) +static inline +void put_futex_key(int fshared, union futex_key *key) { - if (key->both.ptr != 0) { - if (key->both.offset & 1) - atomic_inc(&key->shared.inode->i_count); - else - atomic_inc(&key->private.mm->mm_count); - } + drop_futex_key_refs(key); } -/* - * Drop a reference to the resource addressed by a key. - * The hash bucket spinlock must not be held. - */ -static void drop_key_refs(union futex_key *key) +static u32 cmpxchg_futex_value_locked(u32 __user *uaddr, u32 uval, u32 newval) { - if (key->both.ptr != 0) { - if (key->both.offset & 1) - iput(key->shared.inode); - else - mmdrop(key->private.mm); - } + u32 curval; + + pagefault_disable(); + curval = futex_atomic_cmpxchg_inatomic(uaddr, uval, newval); + pagefault_enable(); + + return curval; } -static inline int get_futex_value_locked(u32 *dest, u32 __user *from) +static int get_futex_value_locked(u32 *dest, u32 __user *from) { int ret; - inc_preempt_count(); + pagefault_disable(); ret = __copy_from_user_inatomic(dest, from, sizeof(u32)); - dec_preempt_count(); + pagefault_enable(); return ret ? -EFAULT : 0; } /* - * Fault handling. Called with current->mm->mmap_sem held. + * Fault handling. */ static int futex_handle_fault(unsigned long address, int attempt) { struct vm_area_struct * vma; struct mm_struct *mm = current->mm; + int ret = -EFAULT; - if (attempt >= 2 || !(vma = find_vma(mm, address)) || - vma->vm_start > address || !(vma->vm_flags & VM_WRITE)) - return -EFAULT; + if (attempt > 2) + return ret; - switch (handle_mm_fault(mm, vma, address, 1)) { - case VM_FAULT_MINOR: - current->min_flt++; - break; - case VM_FAULT_MAJOR: - current->maj_flt++; - break; - default: - return -EFAULT; + down_read(&mm->mmap_sem); + vma = find_vma(mm, address); + if (vma && address >= vma->vm_start && + (vma->vm_flags & VM_WRITE)) { + int fault; + fault = handle_mm_fault(mm, vma, address, 1); + if (unlikely((fault & VM_FAULT_ERROR))) { +#if 0 + /* XXX: let's do this when we verify it is OK */ + if (ret & VM_FAULT_OOM) + ret = -ENOMEM; +#endif + } else { + ret = 0; + if (fault & VM_FAULT_MAJOR) + current->maj_flt++; + else + current->min_flt++; + } } - return 0; + up_read(&mm->mmap_sem); + return ret; } /* @@ -324,16 +345,16 @@ static int refill_pi_state_cache(void) if (likely(current->pi_state_cache)) return 0; - pi_state = kmalloc(sizeof(*pi_state), GFP_KERNEL); + pi_state = kzalloc(sizeof(*pi_state), GFP_KERNEL); if (!pi_state) return -ENOMEM; - memset(pi_state, 0, sizeof(*pi_state)); INIT_LIST_HEAD(&pi_state->list); /* pi_mutex gets initialized later */ pi_state->owner = NULL; atomic_set(&pi_state->refcount, 1); + pi_state->key = FUTEX_KEY_INIT; current->pi_state_cache = pi_state; @@ -388,22 +409,22 @@ static void free_pi_state(struct futex_pi_state *pi_state) static struct task_struct * futex_find_get_task(pid_t pid) { struct task_struct *p; + const struct cred *cred = current_cred(), *pcred; - read_lock(&tasklist_lock); - p = find_task_by_pid(pid); - if (!p) - goto out_unlock; - if ((current->euid != p->euid) && (current->euid != p->uid)) { - p = NULL; - goto out_unlock; - } - if (p->state == EXIT_ZOMBIE || p->exit_state == EXIT_ZOMBIE) { - p = NULL; - goto out_unlock; + rcu_read_lock(); + p = find_task_by_vpid(pid); + if (!p) { + p = ERR_PTR(-ESRCH); + } else { + pcred = __task_cred(p); + if (cred->euid != pcred->euid && + cred->euid != pcred->uid) + p = ERR_PTR(-ESRCH); + else + get_task_struct(p); } - get_task_struct(p); -out_unlock: - read_unlock(&tasklist_lock); + + rcu_read_unlock(); return p; } @@ -415,15 +436,17 @@ out_unlock: */ void exit_pi_state_list(struct task_struct *curr) { - struct futex_hash_bucket *hb; struct list_head *next, *head = &curr->pi_state_list; struct futex_pi_state *pi_state; - union futex_key key; + struct futex_hash_bucket *hb; + union futex_key key = FUTEX_KEY_INIT; + if (!futex_cmpxchg_enabled) + return; /* * We are a ZOMBIE and nobody can enqueue itself on * pi_state_list anymore, but we have to be careful - * versus waiters unqueueing themselfs + * versus waiters unqueueing themselves: */ spin_lock_irq(&curr->pi_lock); while (!list_empty(head)) { @@ -431,21 +454,24 @@ void exit_pi_state_list(struct task_struct *curr) next = head->next; pi_state = list_entry(next, struct futex_pi_state, list); key = pi_state->key; + hb = hash_futex(&key); spin_unlock_irq(&curr->pi_lock); - hb = hash_futex(&key); spin_lock(&hb->lock); spin_lock_irq(&curr->pi_lock); + /* + * We dropped the pi-lock, so re-check whether this + * task still owns the PI-state: + */ if (head->next != next) { spin_unlock(&hb->lock); continue; } - list_del_init(&pi_state->list); - WARN_ON(pi_state->owner != curr); - + WARN_ON(list_empty(&pi_state->list)); + list_del_init(&pi_state->list); pi_state->owner = NULL; spin_unlock_irq(&curr->pi_lock); @@ -459,18 +485,19 @@ void exit_pi_state_list(struct task_struct *curr) } static int -lookup_pi_state(u32 uval, struct futex_hash_bucket *hb, struct futex_q *me) +lookup_pi_state(u32 uval, struct futex_hash_bucket *hb, + union futex_key *key, struct futex_pi_state **ps) { struct futex_pi_state *pi_state = NULL; struct futex_q *this, *next; - struct list_head *head; + struct plist_head *head; struct task_struct *p; - pid_t pid; + pid_t pid = uval & FUTEX_TID_MASK; head = &hb->chain; - list_for_each_entry_safe(this, next, head, list) { - if (match_futex (&this->key, &me->key)) { + plist_for_each_entry_safe(this, next, head, list) { + if (match_futex(&this->key, key)) { /* * Another waiter already exists - bump up * the refcount and return its pi_state: @@ -482,21 +509,46 @@ lookup_pi_state(u32 uval, struct futex_hash_bucket *hb, struct futex_q *me) if (unlikely(!pi_state)) return -EINVAL; + WARN_ON(!atomic_read(&pi_state->refcount)); + WARN_ON(pid && pi_state->owner && + pi_state->owner->pid != pid); + atomic_inc(&pi_state->refcount); - me->pi_state = pi_state; + *ps = pi_state; return 0; } } /* - * We are the first waiter - try to look up the real owner and - * attach the new pi_state to it: + * We are the first waiter - try to look up the real owner and attach + * the new pi_state to it, but bail out when TID = 0 */ - pid = uval & FUTEX_TID_MASK; - p = futex_find_get_task(pid); - if (!p) + if (!pid) return -ESRCH; + p = futex_find_get_task(pid); + if (IS_ERR(p)) + return PTR_ERR(p); + + /* + * We need to look at the task state flags to figure out, + * whether the task is exiting. To protect against the do_exit + * change of the task flags, we do this protected by + * p->pi_lock: + */ + spin_lock_irq(&p->pi_lock); + if (unlikely(p->flags & PF_EXITING)) { + /* + * The task is on the way out. When PF_EXITPIDONE is + * set, we know that the task has finished the + * cleanup: + */ + int ret = (p->flags & PF_EXITPIDONE) ? -ESRCH : -EAGAIN; + + spin_unlock_irq(&p->pi_lock); + put_task_struct(p); + return ret; + } pi_state = alloc_pi_state(); @@ -507,16 +559,16 @@ lookup_pi_state(u32 uval, struct futex_hash_bucket *hb, struct futex_q *me) rt_mutex_init_proxy_locked(&pi_state->pi_mutex, p); /* Store the key for possible exit cleanups: */ - pi_state->key = me->key; + pi_state->key = *key; - spin_lock_irq(&p->pi_lock); + WARN_ON(!list_empty(&pi_state->list)); list_add(&pi_state->list, &p->pi_state_list); pi_state->owner = p; spin_unlock_irq(&p->pi_lock); put_task_struct(p); - me->pi_state = pi_state; + *ps = pi_state; return 0; } @@ -527,14 +579,12 @@ lookup_pi_state(u32 uval, struct futex_hash_bucket *hb, struct futex_q *me) */ static void wake_futex(struct futex_q *q) { - list_del_init(&q->list); - if (q->filp) - send_sigio(&q->filp->f_owner, q->fd, POLL_IN); + plist_del(&q->list, &q->list.plist); /* * The lock in wake_up_all() is a crucial memory barrier after the - * list_del_init() and also before assigning to q->lock_ptr. + * plist_del() and also before assigning to q->lock_ptr. */ - wake_up_all(&q->waiters); + wake_up(&q->waiter); /* * The waiting task can free the futex_q as soon as this is written, * without taking any locks. This must come last. @@ -544,7 +594,7 @@ static void wake_futex(struct futex_q *q) * at the end of wake_up_all() does not prevent this store from * moving. */ - wmb(); + smp_wmb(); q->lock_ptr = NULL; } @@ -557,6 +607,7 @@ static int wake_futex_pi(u32 __user *uaddr, u32 uval, struct futex_q *this) if (!pi_state) return -EINVAL; + spin_lock(&pi_state->pi_mutex.wait_lock); new_owner = rt_mutex_next_owner(&pi_state->pi_mutex); /* @@ -573,20 +624,35 @@ static int wake_futex_pi(u32 __user *uaddr, u32 uval, struct futex_q *this) * kept enabled while there is PI state around. We must also * preserve the owner died bit.) */ - newval = (uval & FUTEX_OWNER_DIED) | FUTEX_WAITERS | new_owner->pid; + if (!(uval & FUTEX_OWNER_DIED)) { + int ret = 0; - inc_preempt_count(); - curval = futex_atomic_cmpxchg_inatomic(uaddr, uval, newval); - dec_preempt_count(); + newval = FUTEX_WAITERS | task_pid_vnr(new_owner); - if (curval == -EFAULT) - return -EFAULT; - if (curval != uval) - return -EINVAL; + curval = cmpxchg_futex_value_locked(uaddr, uval, newval); + + if (curval == -EFAULT) + ret = -EFAULT; + else if (curval != uval) + ret = -EINVAL; + if (ret) { + spin_unlock(&pi_state->pi_mutex.wait_lock); + return ret; + } + } + + spin_lock_irq(&pi_state->owner->pi_lock); + WARN_ON(list_empty(&pi_state->list)); + list_del_init(&pi_state->list); + spin_unlock_irq(&pi_state->owner->pi_lock); - list_del_init(&pi_state->owner->pi_state_list); + spin_lock_irq(&new_owner->pi_lock); + WARN_ON(!list_empty(&pi_state->list)); list_add(&pi_state->list, &new_owner->pi_state_list); pi_state->owner = new_owner; + spin_unlock_irq(&new_owner->pi_lock); + + spin_unlock(&pi_state->pi_mutex.wait_lock); rt_mutex_unlock(&pi_state->pi_mutex); return 0; @@ -600,9 +666,7 @@ static int unlock_futex_pi(u32 __user *uaddr, u32 uval) * There is no waiter, so we unlock the futex. The owner died * bit has not to be preserved here. We are the owner: */ - inc_preempt_count(); - oldval = futex_atomic_cmpxchg_inatomic(uaddr, uval, 0); - dec_preempt_count(); + oldval = cmpxchg_futex_value_locked(uaddr, uval, 0); if (oldval == -EFAULT) return oldval; @@ -632,17 +696,18 @@ double_lock_hb(struct futex_hash_bucket *hb1, struct futex_hash_bucket *hb2) * Wake up all waiters hashed on the physical page that is mapped * to this virtual address: */ -static int futex_wake(u32 __user *uaddr, int nr_wake) +static int futex_wake(u32 __user *uaddr, int fshared, int nr_wake, u32 bitset) { struct futex_hash_bucket *hb; struct futex_q *this, *next; - struct list_head *head; - union futex_key key; + struct plist_head *head; + union futex_key key = FUTEX_KEY_INIT; int ret; - down_read(¤t->mm->mmap_sem); + if (!bitset) + return -EINVAL; - ret = get_futex_key(uaddr, &key); + ret = get_futex_key(uaddr, fshared, &key); if (unlikely(ret != 0)) goto out; @@ -650,12 +715,17 @@ static int futex_wake(u32 __user *uaddr, int nr_wake) spin_lock(&hb->lock); head = &hb->chain; - list_for_each_entry_safe(this, next, head, list) { + plist_for_each_entry_safe(this, next, head, list) { if (match_futex (&this->key, &key)) { if (this->pi_state) { ret = -EINVAL; break; } + + /* Check if one of the bits is set in both bitsets */ + if (!(this->bitset & bitset)) + continue; + wake_futex(this); if (++ret >= nr_wake) break; @@ -663,8 +733,8 @@ static int futex_wake(u32 __user *uaddr, int nr_wake) } spin_unlock(&hb->lock); + put_futex_key(fshared, &key); out: - up_read(¤t->mm->mmap_sem); return ret; } @@ -673,24 +743,22 @@ out: * to this virtual address: */ static int -futex_wake_op(u32 __user *uaddr1, u32 __user *uaddr2, +futex_wake_op(u32 __user *uaddr1, int fshared, u32 __user *uaddr2, int nr_wake, int nr_wake2, int op) { - union futex_key key1, key2; + union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT; struct futex_hash_bucket *hb1, *hb2; - struct list_head *head; + struct plist_head *head; struct futex_q *this, *next; int ret, op_ret, attempt = 0; retryfull: - down_read(¤t->mm->mmap_sem); - - ret = get_futex_key(uaddr1, &key1); + ret = get_futex_key(uaddr1, fshared, &key1); if (unlikely(ret != 0)) goto out; - ret = get_futex_key(uaddr2, &key2); + ret = get_futex_key(uaddr2, fshared, &key2); if (unlikely(ret != 0)) - goto out; + goto out_put_key1; hb1 = hash_futex(&key1); hb2 = hash_futex(&key2); @@ -712,12 +780,12 @@ retry: * but we might get them from range checking */ ret = op_ret; - goto out; + goto out_put_keys; #endif if (unlikely(op_ret != -EFAULT)) { ret = op_ret; - goto out; + goto out_put_keys; } /* @@ -728,18 +796,13 @@ retry: * still holding the mmap_sem. */ if (attempt++) { - if (futex_handle_fault((unsigned long)uaddr2, - attempt)) - goto out; + ret = futex_handle_fault((unsigned long)uaddr2, + attempt); + if (ret) + goto out_put_keys; goto retry; } - /* - * If we would have faulted, release mmap_sem, - * fault it in and start all over again. - */ - up_read(¤t->mm->mmap_sem); - ret = get_user(dummy, uaddr2); if (ret) return ret; @@ -749,7 +812,7 @@ retry: head = &hb1->chain; - list_for_each_entry_safe(this, next, head, list) { + plist_for_each_entry_safe(this, next, head, list) { if (match_futex (&this->key, &key1)) { wake_futex(this); if (++ret >= nr_wake) @@ -761,7 +824,7 @@ retry: head = &hb2->chain; op_ret = 0; - list_for_each_entry_safe(this, next, head, list) { + plist_for_each_entry_safe(this, next, head, list) { if (match_futex (&this->key, &key2)) { wake_futex(this); if (++op_ret >= nr_wake2) @@ -774,8 +837,11 @@ retry: spin_unlock(&hb1->lock); if (hb1 != hb2) spin_unlock(&hb2->lock); +out_put_keys: + put_futex_key(fshared, &key2); +out_put_key1: + put_futex_key(fshared, &key1); out: - up_read(¤t->mm->mmap_sem); return ret; } @@ -783,24 +849,22 @@ out: * Requeue all waiters hashed on one physical page to another * physical page. */ -static int futex_requeue(u32 __user *uaddr1, u32 __user *uaddr2, +static int futex_requeue(u32 __user *uaddr1, int fshared, u32 __user *uaddr2, int nr_wake, int nr_requeue, u32 *cmpval) { - union futex_key key1, key2; + union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT; struct futex_hash_bucket *hb1, *hb2; - struct list_head *head1; + struct plist_head *head1; struct futex_q *this, *next; int ret, drop_count = 0; - retry: - down_read(¤t->mm->mmap_sem); - - ret = get_futex_key(uaddr1, &key1); +retry: + ret = get_futex_key(uaddr1, fshared, &key1); if (unlikely(ret != 0)) goto out; - ret = get_futex_key(uaddr2, &key2); + ret = get_futex_key(uaddr2, fshared, &key2); if (unlikely(ret != 0)) - goto out; + goto out_put_key1; hb1 = hash_futex(&key1); hb2 = hash_futex(&key2); @@ -817,18 +881,12 @@ static int futex_requeue(u32 __user *uaddr1, u32 __user *uaddr2, if (hb1 != hb2) spin_unlock(&hb2->lock); - /* - * If we would have faulted, release mmap_sem, fault - * it in and start all over again. - */ - up_read(¤t->mm->mmap_sem); - ret = get_user(curval, uaddr1); if (!ret) goto retry; - return ret; + goto out_put_keys; } if (curval != *cmpval) { ret = -EAGAIN; @@ -837,7 +895,7 @@ static int futex_requeue(u32 __user *uaddr1, u32 __user *uaddr2, } head1 = &hb1->chain; - list_for_each_entry_safe(this, next, head1, list) { + plist_for_each_entry_safe(this, next, head1, list) { if (!match_futex (&this->key, &key1)) continue; if (++ret <= nr_wake) { @@ -848,11 +906,15 @@ static int futex_requeue(u32 __user *uaddr1, u32 __user *uaddr2, * requeue. */ if (likely(head1 != &hb2->chain)) { - list_move_tail(&this->list, &hb2->chain); + plist_del(&this->list, &hb1->chain); + plist_add(&this->list, &hb2->chain); this->lock_ptr = &hb2->lock; +#ifdef CONFIG_DEBUG_PI_LIST + this->list.plist.lock = &hb2->lock; +#endif } this->key = key2; - get_key_refs(&key2); + get_futex_key_refs(&key2); drop_count++; if (ret - nr_wake >= nr_requeue) @@ -865,27 +927,26 @@ out_unlock: if (hb1 != hb2) spin_unlock(&hb2->lock); - /* drop_key_refs() must be called outside the spinlocks. */ + /* drop_futex_key_refs() must be called outside the spinlocks. */ while (--drop_count >= 0) - drop_key_refs(&key1); + drop_futex_key_refs(&key1); +out_put_keys: + put_futex_key(fshared, &key2); +out_put_key1: + put_futex_key(fshared, &key1); out: - up_read(¤t->mm->mmap_sem); return ret; } /* The key must be already stored in q->key. */ -static inline struct futex_hash_bucket * -queue_lock(struct futex_q *q, int fd, struct file *filp) +static inline struct futex_hash_bucket *queue_lock(struct futex_q *q) { struct futex_hash_bucket *hb; - q->fd = fd; - q->filp = filp; - - init_waitqueue_head(&q->waiters); + init_waitqueue_head(&q->waiter); - get_key_refs(&q->key); + get_futex_key_refs(&q->key); hb = hash_futex(&q->key); q->lock_ptr = &hb->lock; @@ -893,9 +954,25 @@ queue_lock(struct futex_q *q, int fd, struct file *filp) return hb; } -static inline void __queue_me(struct futex_q *q, struct futex_hash_bucket *hb) +static inline void queue_me(struct futex_q *q, struct futex_hash_bucket *hb) { - list_add_tail(&q->list, &hb->chain); + int prio; + + /* + * The priority used to register this element is + * - either the real thread-priority for the real-time threads + * (i.e. threads with a priority lower than MAX_RT_PRIO) + * - or MAX_RT_PRIO for non-RT threads. + * Thus, all RT-threads are woken first in priority order, and + * the others are woken last, in FIFO order. + */ + prio = min(current->normal_prio, MAX_RT_PRIO); + + plist_node_init(&q->list, prio); +#ifdef CONFIG_DEBUG_PI_LIST + q->list.plist.lock = &hb->lock; +#endif + plist_add(&q->list, &hb->chain); q->task = current; spin_unlock(&hb->lock); } @@ -904,7 +981,7 @@ static inline void queue_unlock(struct futex_q *q, struct futex_hash_bucket *hb) { spin_unlock(&hb->lock); - drop_key_refs(&q->key); + drop_futex_key_refs(&q->key); } /* @@ -912,15 +989,6 @@ queue_unlock(struct futex_q *q, struct futex_hash_bucket *hb) * exactly once. They are called with the hashed spinlock held. */ -/* The key must be already stored in q->key. */ -static void queue_me(struct futex_q *q, int fd, struct file *filp) -{ - struct futex_hash_bucket *hb; - - hb = queue_lock(q, fd, filp); - __queue_me(q, hb); -} - /* Return 1 if we were still queued (ie. 0 means we were woken) */ static int unqueue_me(struct futex_q *q) { @@ -928,9 +996,10 @@ static int unqueue_me(struct futex_q *q) int ret = 0; /* In the common case we don't take the spinlock, which is nice. */ - retry: +retry: lock_ptr = q->lock_ptr; - if (lock_ptr != 0) { + barrier(); + if (lock_ptr != NULL) { spin_lock(lock_ptr); /* * q->lock_ptr can change between reading it and @@ -949,8 +1018,8 @@ static int unqueue_me(struct futex_q *q) spin_unlock(lock_ptr); goto retry; } - WARN_ON(list_empty(&q->list)); - list_del(&q->list); + WARN_ON(plist_node_empty(&q->list)); + plist_del(&q->list, &q->list.plist); BUG_ON(q->pi_state); @@ -958,46 +1027,165 @@ static int unqueue_me(struct futex_q *q) ret = 1; } - drop_key_refs(&q->key); + drop_futex_key_refs(&q->key); return ret; } /* * PI futexes can not be requeued and must remove themself from the - * hash bucket. The hash bucket lock is held on entry and dropped here. + * hash bucket. The hash bucket lock (i.e. lock_ptr) is held on entry + * and dropped here. */ -static void unqueue_me_pi(struct futex_q *q, struct futex_hash_bucket *hb) +static void unqueue_me_pi(struct futex_q *q) { - WARN_ON(list_empty(&q->list)); - list_del(&q->list); + WARN_ON(plist_node_empty(&q->list)); + plist_del(&q->list, &q->list.plist); BUG_ON(!q->pi_state); free_pi_state(q->pi_state); q->pi_state = NULL; - spin_unlock(&hb->lock); + spin_unlock(q->lock_ptr); - drop_key_refs(&q->key); + drop_futex_key_refs(&q->key); } -static int futex_wait(u32 __user *uaddr, u32 val, unsigned long time) +/* + * Fixup the pi_state owner with the new owner. + * + * Must be called with hash bucket lock held and mm->sem held for non + * private futexes. + */ +static int fixup_pi_state_owner(u32 __user *uaddr, struct futex_q *q, + struct task_struct *newowner, int fshared) +{ + u32 newtid = task_pid_vnr(newowner) | FUTEX_WAITERS; + struct futex_pi_state *pi_state = q->pi_state; + struct task_struct *oldowner = pi_state->owner; + u32 uval, curval, newval; + int ret, attempt = 0; + + /* Owner died? */ + if (!pi_state->owner) + newtid |= FUTEX_OWNER_DIED; + + /* + * We are here either because we stole the rtmutex from the + * pending owner or we are the pending owner which failed to + * get the rtmutex. We have to replace the pending owner TID + * in the user space variable. This must be atomic as we have + * to preserve the owner died bit here. + * + * Note: We write the user space value _before_ changing the + * pi_state because we can fault here. Imagine swapped out + * pages or a fork, which was running right before we acquired + * mmap_sem, that marked all the anonymous memory readonly for + * cow. + * + * Modifying pi_state _before_ the user space value would + * leave the pi_state in an inconsistent state when we fault + * here, because we need to drop the hash bucket lock to + * handle the fault. This might be observed in the PID check + * in lookup_pi_state. + */ +retry: + if (get_futex_value_locked(&uval, uaddr)) + goto handle_fault; + + while (1) { + newval = (uval & FUTEX_OWNER_DIED) | newtid; + + curval = cmpxchg_futex_value_locked(uaddr, uval, newval); + + if (curval == -EFAULT) + goto handle_fault; + if (curval == uval) + break; + uval = curval; + } + + /* + * We fixed up user space. Now we need to fix the pi_state + * itself. + */ + if (pi_state->owner != NULL) { + spin_lock_irq(&pi_state->owner->pi_lock); + WARN_ON(list_empty(&pi_state->list)); + list_del_init(&pi_state->list); + spin_unlock_irq(&pi_state->owner->pi_lock); + } + + pi_state->owner = newowner; + + spin_lock_irq(&newowner->pi_lock); + WARN_ON(!list_empty(&pi_state->list)); + list_add(&pi_state->list, &newowner->pi_state_list); + spin_unlock_irq(&newowner->pi_lock); + return 0; + + /* + * To handle the page fault we need to drop the hash bucket + * lock here. That gives the other task (either the pending + * owner itself or the task which stole the rtmutex) the + * chance to try the fixup of the pi_state. So once we are + * back from handling the fault we need to check the pi_state + * after reacquiring the hash bucket lock and before trying to + * do another fixup. When the fixup has been done already we + * simply return. + */ +handle_fault: + spin_unlock(q->lock_ptr); + + ret = futex_handle_fault((unsigned long)uaddr, attempt++); + + spin_lock(q->lock_ptr); + + /* + * Check if someone else fixed it for us: + */ + if (pi_state->owner != oldowner) + return 0; + + if (ret) + return ret; + + goto retry; +} + +/* + * In case we must use restart_block to restart a futex_wait, + * we encode in the 'flags' shared capability + */ +#define FLAGS_SHARED 0x01 +#define FLAGS_CLOCKRT 0x02 + +static long futex_wait_restart(struct restart_block *restart); + +static int futex_wait(u32 __user *uaddr, int fshared, + u32 val, ktime_t *abs_time, u32 bitset, int clockrt) { struct task_struct *curr = current; + struct restart_block *restart; DECLARE_WAITQUEUE(wait, curr); struct futex_hash_bucket *hb; struct futex_q q; u32 uval; int ret; + struct hrtimer_sleeper t; + int rem = 0; - q.pi_state = NULL; - retry: - down_read(&curr->mm->mmap_sem); + if (!bitset) + return -EINVAL; - ret = get_futex_key(uaddr, &q.key); + q.pi_state = NULL; + q.bitset = bitset; +retry: + q.key = FUTEX_KEY_INIT; + ret = get_futex_key(uaddr, fshared, &q.key); if (unlikely(ret != 0)) - goto out_release_sem; + goto out; - hb = queue_lock(&q, -1, NULL); + hb = queue_lock(&q); /* * Access the page AFTER the futex is queued. @@ -1016,38 +1204,29 @@ static int futex_wait(u32 __user *uaddr, u32 val, unsigned long time) * a wakeup when *uaddr != val on entry to the syscall. This is * rare, but normal. * - * We hold the mmap semaphore, so the mapping cannot have changed - * since we looked it up in get_futex_key. + * for shared futexes, we hold the mmap semaphore, so the mapping + * cannot have changed since we looked it up in get_futex_key. */ ret = get_futex_value_locked(&uval, uaddr); if (unlikely(ret)) { queue_unlock(&q, hb); - - /* - * If we would have faulted, release mmap_sem, fault it in and - * start all over again. - */ - up_read(&curr->mm->mmap_sem); + put_futex_key(fshared, &q.key); ret = get_user(uval, uaddr); if (!ret) goto retry; - return ret; + goto out; } ret = -EWOULDBLOCK; - if (uval != val) - goto out_unlock_release_sem; + if (unlikely(uval != val)) { + queue_unlock(&q, hb); + goto out_put_key; + } /* Only actually queue if *uaddr contained val. */ - __queue_me(&q, hb); - - /* - * Now the futex is queued and we have checked the data, we - * don't want to hold mmap_sem while we sleep. - */ - up_read(&curr->mm->mmap_sem); + queue_me(&q, hb); /* * There might have been scheduling since the queue_me(), as we @@ -1060,13 +1239,46 @@ static int futex_wait(u32 __user *uaddr, u32 val, unsigned long time) /* add_wait_queue is the barrier after __set_current_state. */ __set_current_state(TASK_INTERRUPTIBLE); - add_wait_queue(&q.waiters, &wait); + add_wait_queue(&q.waiter, &wait); /* - * !list_empty() is safe here without any lock. + * !plist_node_empty() is safe here without any lock. * q.lock_ptr != 0 is not safe, because of ordering against wakeup. */ - if (likely(!list_empty(&q.list))) - time = schedule_timeout(time); + if (likely(!plist_node_empty(&q.list))) { + if (!abs_time) + schedule(); + else { + unsigned long slack; + slack = current->timer_slack_ns; + if (rt_task(current)) + slack = 0; + hrtimer_init_on_stack(&t.timer, + clockrt ? CLOCK_REALTIME : + CLOCK_MONOTONIC, + HRTIMER_MODE_ABS); + hrtimer_init_sleeper(&t, current); + hrtimer_set_expires_range_ns(&t.timer, *abs_time, slack); + + hrtimer_start_expires(&t.timer, HRTIMER_MODE_ABS); + if (!hrtimer_active(&t.timer)) + t.task = NULL; + + /* + * the timer could have already expired, in which + * case current would be flagged for rescheduling. + * Don't bother calling schedule. + */ + if (likely(t.task)) + schedule(); + + hrtimer_cancel(&t.timer); + + /* Flag if a timeout occured */ + rem = (t.task == NULL); + + destroy_hrtimer_on_stack(&t.timer); + } + } __set_current_state(TASK_RUNNING); /* @@ -1075,88 +1287,150 @@ static int futex_wait(u32 __user *uaddr, u32 val, unsigned long time) */ /* If we were woken (and unqueued), we succeeded, whatever. */ + ret = 0; if (!unqueue_me(&q)) - return 0; - if (time == 0) - return -ETIMEDOUT; + goto out_put_key; + ret = -ETIMEDOUT; + if (rem) + goto out_put_key; + /* * We expect signal_pending(current), but another thread may * have handled it for us already. */ - return -EINTR; + ret = -ERESTARTSYS; + if (!abs_time) + goto out_put_key; - out_unlock_release_sem: - queue_unlock(&q, hb); - - out_release_sem: - up_read(&curr->mm->mmap_sem); + restart = ¤t_thread_info()->restart_block; + restart->fn = futex_wait_restart; + restart->futex.uaddr = (u32 *)uaddr; + restart->futex.val = val; + restart->futex.time = abs_time->tv64; + restart->futex.bitset = bitset; + restart->futex.flags = 0; + + if (fshared) + restart->futex.flags |= FLAGS_SHARED; + if (clockrt) + restart->futex.flags |= FLAGS_CLOCKRT; + + ret = -ERESTART_RESTARTBLOCK; + +out_put_key: + put_futex_key(fshared, &q.key); +out: return ret; } + +static long futex_wait_restart(struct restart_block *restart) +{ + u32 __user *uaddr = (u32 __user *)restart->futex.uaddr; + int fshared = 0; + ktime_t t; + + t.tv64 = restart->futex.time; + restart->fn = do_no_restart_syscall; + if (restart->futex.flags & FLAGS_SHARED) + fshared = 1; + return (long)futex_wait(uaddr, fshared, restart->futex.val, &t, + restart->futex.bitset, + restart->futex.flags & FLAGS_CLOCKRT); +} + + /* * Userspace tried a 0 -> TID atomic transition of the futex value * and failed. The kernel side here does the whole locking operation: * if there are waiters then it will block, it does PI, etc. (Due to * races the kernel might see a 0 value of the futex too.) */ -static int do_futex_lock_pi(u32 __user *uaddr, int detect, int trylock, - struct hrtimer_sleeper *to) +static int futex_lock_pi(u32 __user *uaddr, int fshared, + int detect, ktime_t *time, int trylock) { + struct hrtimer_sleeper timeout, *to = NULL; struct task_struct *curr = current; struct futex_hash_bucket *hb; u32 uval, newval, curval; struct futex_q q; - int ret, attempt = 0; + int ret, lock_taken, ownerdied = 0, attempt = 0; if (refill_pi_state_cache()) return -ENOMEM; - q.pi_state = NULL; - retry: - down_read(&curr->mm->mmap_sem); + if (time) { + to = &timeout; + hrtimer_init_on_stack(&to->timer, CLOCK_REALTIME, + HRTIMER_MODE_ABS); + hrtimer_init_sleeper(to, current); + hrtimer_set_expires(&to->timer, *time); + } - ret = get_futex_key(uaddr, &q.key); + q.pi_state = NULL; +retry: + q.key = FUTEX_KEY_INIT; + ret = get_futex_key(uaddr, fshared, &q.key); if (unlikely(ret != 0)) - goto out_release_sem; + goto out; + +retry_unlocked: + hb = queue_lock(&q); - hb = queue_lock(&q, -1, NULL); +retry_locked: + ret = lock_taken = 0; - retry_locked: /* * To avoid races, we attempt to take the lock here again * (by doing a 0 -> TID atomic cmpxchg), while holding all * the locks. It will most likely not succeed. */ - newval = current->pid; + newval = task_pid_vnr(current); - inc_preempt_count(); - curval = futex_atomic_cmpxchg_inatomic(uaddr, 0, newval); - dec_preempt_count(); + curval = cmpxchg_futex_value_locked(uaddr, 0, newval); if (unlikely(curval == -EFAULT)) goto uaddr_faulted; - /* We own the lock already */ - if (unlikely((curval & FUTEX_TID_MASK) == current->pid)) { - if (!detect && 0) - force_sig(SIGKILL, current); + /* + * Detect deadlocks. In case of REQUEUE_PI this is a valid + * situation and we return success to user space. + */ + if (unlikely((curval & FUTEX_TID_MASK) == task_pid_vnr(current))) { ret = -EDEADLK; - goto out_unlock_release_sem; + goto out_unlock_put_key; } /* - * Surprise - we got the lock. Just return - * to userspace: + * Surprise - we got the lock. Just return to userspace: */ if (unlikely(!curval)) - goto out_unlock_release_sem; + goto out_unlock_put_key; uval = curval; - newval = uval | FUTEX_WAITERS; - inc_preempt_count(); - curval = futex_atomic_cmpxchg_inatomic(uaddr, uval, newval); - dec_preempt_count(); + /* + * Set the WAITERS flag, so the owner will know it has someone + * to wake at next unlock + */ + newval = curval | FUTEX_WAITERS; + + /* + * There are two cases, where a futex might have no owner (the + * owner TID is 0): OWNER_DIED. We take over the futex in this + * case. We also do an unconditional take over, when the owner + * of the futex died. + * + * This is safe as we are protected by the hash bucket lock ! + */ + if (unlikely(ownerdied || !(curval & FUTEX_TID_MASK))) { + /* Keep the OWNER_DIED bit */ + newval = (curval & ~FUTEX_TID_MASK) | task_pid_vnr(current); + ownerdied = 0; + lock_taken = 1; + } + + curval = cmpxchg_futex_value_locked(uaddr, uval, newval); if (unlikely(curval == -EFAULT)) goto uaddr_faulted; @@ -1164,52 +1438,56 @@ static int do_futex_lock_pi(u32 __user *uaddr, int detect, int trylock, goto retry_locked; /* + * We took the lock due to owner died take over. + */ + if (unlikely(lock_taken)) + goto out_unlock_put_key; + + /* * We dont have the lock. Look up the PI state (or create it if * we are the first waiter): */ - ret = lookup_pi_state(uval, hb, &q); + ret = lookup_pi_state(uval, hb, &q.key, &q.pi_state); if (unlikely(ret)) { - /* - * There were no waiters and the owner task lookup - * failed. When the OWNER_DIED bit is set, then we - * know that this is a robust futex and we actually - * take the lock. This is safe as we are protected by - * the hash bucket lock. We also set the waiters bit - * unconditionally here, to simplify glibc handling of - * multiple tasks racing to acquire the lock and - * cleanup the problems which were left by the dead - * owner. - */ - if (curval & FUTEX_OWNER_DIED) { - uval = newval; - newval = current->pid | - FUTEX_OWNER_DIED | FUTEX_WAITERS; + switch (ret) { - inc_preempt_count(); - curval = futex_atomic_cmpxchg_inatomic(uaddr, - uval, newval); - dec_preempt_count(); + case -EAGAIN: + /* + * Task is exiting and we just wait for the + * exit to complete. + */ + queue_unlock(&q, hb); + cond_resched(); + goto retry; - if (unlikely(curval == -EFAULT)) + case -ESRCH: + /* + * No owner found for this futex. Check if the + * OWNER_DIED bit is set to figure out whether + * this is a robust futex or not. + */ + if (get_futex_value_locked(&curval, uaddr)) goto uaddr_faulted; - if (unlikely(curval != uval)) + + /* + * We simply start over in case of a robust + * futex. The code above will take the futex + * and return happy. + */ + if (curval & FUTEX_OWNER_DIED) { + ownerdied = 1; goto retry_locked; - ret = 0; + } + default: + goto out_unlock_put_key; } - goto out_unlock_release_sem; } /* * Only actually queue now that the atomic ops are done: */ - __queue_me(&q, hb); - - /* - * Now the futex is queued and we have checked the data, we - * don't want to hold mmap_sem while we sleep. - */ - up_read(&curr->mm->mmap_sem); + queue_me(&q, hb); WARN_ON(!q.pi_state); /* @@ -1223,182 +1501,124 @@ static int do_futex_lock_pi(u32 __user *uaddr, int detect, int trylock, ret = ret ? 0 : -EWOULDBLOCK; } - down_read(&curr->mm->mmap_sem); spin_lock(q.lock_ptr); - /* - * Got the lock. We might not be the anticipated owner if we - * did a lock-steal - fix up the PI-state in that case. - */ - if (!ret && q.pi_state->owner != curr) { - u32 newtid = current->pid | FUTEX_WAITERS; - - /* Owner died? */ - if (q.pi_state->owner != NULL) { - spin_lock_irq(&q.pi_state->owner->pi_lock); - list_del_init(&q.pi_state->list); - spin_unlock_irq(&q.pi_state->owner->pi_lock); - } else - newtid |= FUTEX_OWNER_DIED; - - q.pi_state->owner = current; - - spin_lock_irq(¤t->pi_lock); - list_add(&q.pi_state->list, ¤t->pi_state_list); - spin_unlock_irq(¤t->pi_lock); - - /* Unqueue and drop the lock */ - unqueue_me_pi(&q, hb); - up_read(&curr->mm->mmap_sem); + if (!ret) { /* - * We own it, so we have to replace the pending owner - * TID. This must be atomic as we have preserve the - * owner died bit here. + * Got the lock. We might not be the anticipated owner + * if we did a lock-steal - fix up the PI-state in + * that case: */ - ret = get_user(uval, uaddr); - while (!ret) { - newval = (uval & FUTEX_OWNER_DIED) | newtid; - curval = futex_atomic_cmpxchg_inatomic(uaddr, - uval, newval); - if (curval == -EFAULT) - ret = -EFAULT; - if (curval == uval) - break; - uval = curval; - } + if (q.pi_state->owner != curr) + ret = fixup_pi_state_owner(uaddr, &q, curr, fshared); } else { /* * Catch the rare case, where the lock was released - * when we were on the way back before we locked - * the hash bucket. + * when we were on the way back before we locked the + * hash bucket. */ - if (ret && q.pi_state->owner == curr) { + if (q.pi_state->owner == curr) { + /* + * Try to get the rt_mutex now. This might + * fail as some other task acquired the + * rt_mutex after we removed ourself from the + * rt_mutex waiters list. + */ if (rt_mutex_trylock(&q.pi_state->pi_mutex)) ret = 0; + else { + /* + * pi_state is incorrect, some other + * task did a lock steal and we + * returned due to timeout or signal + * without taking the rt_mutex. Too + * late. We can access the + * rt_mutex_owner without locking, as + * the other task is now blocked on + * the hash bucket lock. Fix the state + * up. + */ + struct task_struct *owner; + int res; + + owner = rt_mutex_owner(&q.pi_state->pi_mutex); + res = fixup_pi_state_owner(uaddr, &q, owner, + fshared); + + /* propagate -EFAULT, if the fixup failed */ + if (res) + ret = res; + } + } else { + /* + * Paranoia check. If we did not take the lock + * in the trylock above, then we should not be + * the owner of the rtmutex, neither the real + * nor the pending one: + */ + if (rt_mutex_owner(&q.pi_state->pi_mutex) == curr) + printk(KERN_ERR "futex_lock_pi: ret = %d " + "pi-mutex: %p pi-state %p\n", ret, + q.pi_state->pi_mutex.owner, + q.pi_state->owner); } - /* Unqueue and drop the lock */ - unqueue_me_pi(&q, hb); - up_read(&curr->mm->mmap_sem); } - if (!detect && ret == -EDEADLK && 0) - force_sig(SIGKILL, current); + /* Unqueue and drop the lock */ + unqueue_me_pi(&q); - return ret; + if (to) + destroy_hrtimer_on_stack(&to->timer); + return ret != -EINTR ? ret : -ERESTARTNOINTR; - out_unlock_release_sem: +out_unlock_put_key: queue_unlock(&q, hb); - out_release_sem: - up_read(&curr->mm->mmap_sem); +out_put_key: + put_futex_key(fshared, &q.key); +out: + if (to) + destroy_hrtimer_on_stack(&to->timer); return ret; - uaddr_faulted: +uaddr_faulted: /* - * We have to r/w *(int __user *)uaddr, but we can't modify it - * non-atomically. Therefore, if get_user below is not - * enough, we need to handle the fault ourselves, while - * still holding the mmap_sem. + * We have to r/w *(int __user *)uaddr, and we have to modify it + * atomically. Therefore, if we continue to fault after get_user() + * below, we need to handle the fault ourselves, while still holding + * the mmap_sem. This can occur if the uaddr is under contention as + * we have to drop the mmap_sem in order to call get_user(). */ - if (attempt++) { - if (futex_handle_fault((unsigned long)uaddr, attempt)) - goto out_unlock_release_sem; + queue_unlock(&q, hb); - goto retry_locked; + if (attempt++) { + ret = futex_handle_fault((unsigned long)uaddr, attempt); + if (ret) + goto out_put_key; + goto retry_unlocked; } - queue_unlock(&q, hb); - up_read(&curr->mm->mmap_sem); - ret = get_user(uval, uaddr); - if (!ret && (uval != -EFAULT)) + if (!ret) goto retry; + if (to) + destroy_hrtimer_on_stack(&to->timer); return ret; } /* - * Restart handler - */ -static long futex_lock_pi_restart(struct restart_block *restart) -{ - struct hrtimer_sleeper timeout, *to = NULL; - int ret; - - restart->fn = do_no_restart_syscall; - - if (restart->arg2 || restart->arg3) { - to = &timeout; - hrtimer_init(&to->timer, CLOCK_REALTIME, HRTIMER_ABS); - hrtimer_init_sleeper(to, current); - to->timer.expires.tv64 = ((u64)restart->arg1 << 32) | - (u64) restart->arg0; - } - - pr_debug("lock_pi restart: %p, %d (%d)\n", - (u32 __user *)restart->arg0, current->pid); - - ret = do_futex_lock_pi((u32 __user *)restart->arg0, restart->arg1, - 0, to); - - if (ret != -EINTR) - return ret; - - restart->fn = futex_lock_pi_restart; - - /* The other values are filled in */ - return -ERESTART_RESTARTBLOCK; -} - -/* - * Called from the syscall entry below. - */ -static int futex_lock_pi(u32 __user *uaddr, int detect, unsigned long sec, - long nsec, int trylock) -{ - struct hrtimer_sleeper timeout, *to = NULL; - struct restart_block *restart; - int ret; - - if (sec != MAX_SCHEDULE_TIMEOUT) { - to = &timeout; - hrtimer_init(&to->timer, CLOCK_REALTIME, HRTIMER_ABS); - hrtimer_init_sleeper(to, current); - to->timer.expires = ktime_set(sec, nsec); - } - - ret = do_futex_lock_pi(uaddr, detect, trylock, to); - - if (ret != -EINTR) - return ret; - - pr_debug("lock_pi interrupted: %p, %d (%d)\n", uaddr, current->pid); - - restart = ¤t_thread_info()->restart_block; - restart->fn = futex_lock_pi_restart; - restart->arg0 = (unsigned long) uaddr; - restart->arg1 = detect; - if (to) { - restart->arg2 = to->timer.expires.tv64 & 0xFFFFFFFF; - restart->arg3 = to->timer.expires.tv64 >> 32; - } else - restart->arg2 = restart->arg3 = 0; - - return -ERESTART_RESTARTBLOCK; -} - -/* * Userspace attempted a TID -> 0 atomic transition, and failed. * This is the in-kernel slowpath: we look up the PI state (if any), * and do the rt-mutex unlock. */ -static int futex_unlock_pi(u32 __user *uaddr) +static int futex_unlock_pi(u32 __user *uaddr, int fshared) { struct futex_hash_bucket *hb; struct futex_q *this, *next; u32 uval; - struct list_head *head; - union futex_key key; + struct plist_head *head; + union futex_key key = FUTEX_KEY_INIT; int ret, attempt = 0; retry: @@ -1407,29 +1627,25 @@ retry: /* * We release only a lock we actually own: */ - if ((uval & FUTEX_TID_MASK) != current->pid) + if ((uval & FUTEX_TID_MASK) != task_pid_vnr(current)) return -EPERM; - /* - * First take all the futex related locks: - */ - down_read(¤t->mm->mmap_sem); - ret = get_futex_key(uaddr, &key); + ret = get_futex_key(uaddr, fshared, &key); if (unlikely(ret != 0)) goto out; hb = hash_futex(&key); +retry_unlocked: spin_lock(&hb->lock); -retry_locked: /* * To avoid races, try to do the TID -> 0 atomic transition * again. If it succeeds then we can return without waking * anyone else up: */ - inc_preempt_count(); - uval = futex_atomic_cmpxchg_inatomic(uaddr, current->pid, 0); - dec_preempt_count(); + if (!(uval & FUTEX_OWNER_DIED)) + uval = cmpxchg_futex_value_locked(uaddr, task_pid_vnr(current), 0); + if (unlikely(uval == -EFAULT)) goto pi_faulted; @@ -1437,7 +1653,7 @@ retry_locked: * Rare case: we managed to release the lock atomically, * no need to wake anyone else up: */ - if (unlikely(uval == current->pid)) + if (unlikely(uval == task_pid_vnr(current))) goto out_unlock; /* @@ -1446,7 +1662,7 @@ retry_locked: */ head = &hb->chain; - list_for_each_entry_safe(this, next, head, list) { + plist_for_each_entry_safe(this, next, head, list) { if (!match_futex (&this->key, &key)) continue; ret = wake_futex_pi(uaddr, uval, this); @@ -1462,147 +1678,44 @@ retry_locked: /* * No waiters - kernel unlocks the futex: */ - ret = unlock_futex_pi(uaddr, uval); - if (ret == -EFAULT) - goto pi_faulted; + if (!(uval & FUTEX_OWNER_DIED)) { + ret = unlock_futex_pi(uaddr, uval); + if (ret == -EFAULT) + goto pi_faulted; + } out_unlock: spin_unlock(&hb->lock); -out: - up_read(¤t->mm->mmap_sem); + put_futex_key(fshared, &key); +out: return ret; pi_faulted: /* - * We have to r/w *(int __user *)uaddr, but we can't modify it - * non-atomically. Therefore, if get_user below is not - * enough, we need to handle the fault ourselves, while - * still holding the mmap_sem. + * We have to r/w *(int __user *)uaddr, and we have to modify it + * atomically. Therefore, if we continue to fault after get_user() + * below, we need to handle the fault ourselves, while still holding + * the mmap_sem. This can occur if the uaddr is under contention as + * we have to drop the mmap_sem in order to call get_user(). */ - if (attempt++) { - if (futex_handle_fault((unsigned long)uaddr, attempt)) - goto out_unlock; + spin_unlock(&hb->lock); - goto retry_locked; + if (attempt++) { + ret = futex_handle_fault((unsigned long)uaddr, attempt); + if (ret) + goto out; + uval = 0; + goto retry_unlocked; } - spin_unlock(&hb->lock); - up_read(¤t->mm->mmap_sem); - ret = get_user(uval, uaddr); - if (!ret && (uval != -EFAULT)) + if (!ret) goto retry; return ret; } -static int futex_close(struct inode *inode, struct file *filp) -{ - struct futex_q *q = filp->private_data; - - unqueue_me(q); - kfree(q); - - return 0; -} - -/* This is one-shot: once it's gone off you need a new fd */ -static unsigned int futex_poll(struct file *filp, - struct poll_table_struct *wait) -{ - struct futex_q *q = filp->private_data; - int ret = 0; - - poll_wait(filp, &q->waiters, wait); - - /* - * list_empty() is safe here without any lock. - * q->lock_ptr != 0 is not safe, because of ordering against wakeup. - */ - if (list_empty(&q->list)) - ret = POLLIN | POLLRDNORM; - - return ret; -} - -static struct file_operations futex_fops = { - .release = futex_close, - .poll = futex_poll, -}; - -/* - * Signal allows caller to avoid the race which would occur if they - * set the sigio stuff up afterwards. - */ -static int futex_fd(u32 __user *uaddr, int signal) -{ - struct futex_q *q; - struct file *filp; - int ret, err; - - ret = -EINVAL; - if (!valid_signal(signal)) - goto out; - - ret = get_unused_fd(); - if (ret < 0) - goto out; - filp = get_empty_filp(); - if (!filp) { - put_unused_fd(ret); - ret = -ENFILE; - goto out; - } - filp->f_op = &futex_fops; - filp->f_vfsmnt = mntget(futex_mnt); - filp->f_dentry = dget(futex_mnt->mnt_root); - filp->f_mapping = filp->f_dentry->d_inode->i_mapping; - - if (signal) { - err = f_setown(filp, current->pid, 1); - if (err < 0) { - goto error; - } - filp->f_owner.signum = signal; - } - - q = kmalloc(sizeof(*q), GFP_KERNEL); - if (!q) { - err = -ENOMEM; - goto error; - } - q->pi_state = NULL; - - down_read(¤t->mm->mmap_sem); - err = get_futex_key(uaddr, &q->key); - - if (unlikely(err != 0)) { - up_read(¤t->mm->mmap_sem); - kfree(q); - goto error; - } - - /* - * queue_me() must be called before releasing mmap_sem, because - * key->shared.inode needs to be referenced while holding it. - */ - filp->private_data = q; - - queue_me(q, ret, filp); - up_read(¤t->mm->mmap_sem); - - /* Now we map fd to filp, so userspace can access it */ - fd_install(ret, filp); -out: - return ret; -error: - put_unused_fd(ret); - put_filp(filp); - ret = err; - goto out; -} - /* * Support for robust futexes: the kernel cleans up held futexes at * thread exit time. @@ -1623,10 +1736,11 @@ error: * @head: pointer to the list-head * @len: length of the list-head, as userspace expects */ -asmlinkage long -sys_set_robust_list(struct robust_list_head __user *head, - size_t len) +SYSCALL_DEFINE2(set_robust_list, struct robust_list_head __user *, head, + size_t, len) { + if (!futex_cmpxchg_enabled) + return -ENOSYS; /* * The kernel knows only one size for now: */ @@ -1644,12 +1758,16 @@ sys_set_robust_list(struct robust_list_head __user *head, * @head_ptr: pointer to a list-head pointer, the kernel fills it in * @len_ptr: pointer to a length field, the kernel fills in the header size */ -asmlinkage long -sys_get_robust_list(int pid, struct robust_list_head __user **head_ptr, - size_t __user *len_ptr) +SYSCALL_DEFINE3(get_robust_list, int, pid, + struct robust_list_head __user * __user *, head_ptr, + size_t __user *, len_ptr) { - struct robust_list_head *head; + struct robust_list_head __user *head; unsigned long ret; + const struct cred *cred = current_cred(), *pcred; + + if (!futex_cmpxchg_enabled) + return -ENOSYS; if (!pid) head = current->robust_list; @@ -1657,16 +1775,18 @@ sys_get_robust_list(int pid, struct robust_list_head __user **head_ptr, struct task_struct *p; ret = -ESRCH; - read_lock(&tasklist_lock); - p = find_task_by_pid(pid); + rcu_read_lock(); + p = find_task_by_vpid(pid); if (!p) goto err_unlock; ret = -EPERM; - if ((current->euid != p->euid) && (current->euid != p->uid) && - !capable(CAP_SYS_PTRACE)) + pcred = __task_cred(p); + if (cred->euid != pcred->euid && + cred->euid != pcred->uid && + !capable(CAP_SYS_PTRACE)) goto err_unlock; head = p->robust_list; - read_unlock(&tasklist_lock); + rcu_read_unlock(); } if (put_user(sizeof(*head), len_ptr)) @@ -1674,7 +1794,7 @@ sys_get_robust_list(int pid, struct robust_list_head __user **head_ptr, return put_user(head, head_ptr); err_unlock: - read_unlock(&tasklist_lock); + rcu_read_unlock(); return ret; } @@ -1683,15 +1803,15 @@ err_unlock: * Process a futex-list entry, check whether it's owned by the * dying task, and do notification if so: */ -int handle_futex_death(u32 __user *uaddr, struct task_struct *curr) +int handle_futex_death(u32 __user *uaddr, struct task_struct *curr, int pi) { - u32 uval, nval; + u32 uval, nval, mval; retry: if (get_user(uval, uaddr)) return -1; - if ((uval & FUTEX_TID_MASK) == curr->pid) { + if ((uval & FUTEX_TID_MASK) == task_pid_vnr(curr)) { /* * Ok, this dying thread is truly holding a futex * of interest. Set the OWNER_DIED bit atomically @@ -1702,21 +1822,44 @@ retry: * thread-death.) The rest of the cleanup is done in * userspace. */ - nval = futex_atomic_cmpxchg_inatomic(uaddr, uval, - uval | FUTEX_OWNER_DIED); + mval = (uval & FUTEX_WAITERS) | FUTEX_OWNER_DIED; + nval = futex_atomic_cmpxchg_inatomic(uaddr, uval, mval); + if (nval == -EFAULT) return -1; if (nval != uval) goto retry; - if (uval & FUTEX_WAITERS) - futex_wake(uaddr, 1); + /* + * Wake robust non-PI futexes here. The wakeup of + * PI futexes happens in exit_pi_state(): + */ + if (!pi && (uval & FUTEX_WAITERS)) + futex_wake(uaddr, 1, 1, FUTEX_BITSET_MATCH_ANY); } return 0; } /* + * Fetch a robust-list pointer. Bit 0 signals PI futexes: + */ +static inline int fetch_robust_entry(struct robust_list __user **entry, + struct robust_list __user * __user *head, + int *pi) +{ + unsigned long uentry; + + if (get_user(uentry, (unsigned long __user *)head)) + return -EFAULT; + + *entry = (void __user *)(uentry & ~1UL); + *pi = uentry & 1; + + return 0; +} + +/* * Walk curr->robust_list (very carefully, it's a userspace list!) * and mark any locks found there dead, and notify any waiters. * @@ -1725,15 +1868,19 @@ retry: void exit_robust_list(struct task_struct *curr) { struct robust_list_head __user *head = curr->robust_list; - struct robust_list __user *entry, *pending; - unsigned int limit = ROBUST_LIST_LIMIT; + struct robust_list __user *entry, *next_entry, *pending; + unsigned int limit = ROBUST_LIST_LIMIT, pi, next_pi, pip; unsigned long futex_offset; + int rc; + + if (!futex_cmpxchg_enabled) + return; /* * Fetch the list head (which was registered earlier, via * sys_set_robust_list()): */ - if (get_user(entry, &head->list.next)) + if (fetch_robust_entry(&entry, &head->list.next, &pi)) return; /* * Fetch the relative futex offset: @@ -1744,25 +1891,28 @@ void exit_robust_list(struct task_struct *curr) * Fetch any possibly pending lock-add first, and handle it * if it exists: */ - if (get_user(pending, &head->list_op_pending)) + if (fetch_robust_entry(&pending, &head->list_op_pending, &pip)) return; - if (pending) - handle_futex_death((void *)pending + futex_offset, curr); + next_entry = NULL; /* avoid warning with gcc */ while (entry != &head->list) { /* + * Fetch the next entry in the list before calling + * handle_futex_death: + */ + rc = fetch_robust_entry(&next_entry, &entry->next, &next_pi); + /* * A pending lock might already be on the list, so * don't process it twice: */ if (entry != pending) - if (handle_futex_death((void *)entry + futex_offset, - curr)) + if (handle_futex_death((void __user *)entry + futex_offset, + curr, pi)) return; - /* - * Fetch the next entry in the list: - */ - if (get_user(entry, &entry->next)) + if (rc) return; + entry = next_entry; + pi = next_pi; /* * Avoid excessively long or circular lists: */ @@ -1771,41 +1921,57 @@ void exit_robust_list(struct task_struct *curr) cond_resched(); } + + if (pending) + handle_futex_death((void __user *)pending + futex_offset, + curr, pip); } -long do_futex(u32 __user *uaddr, int op, u32 val, unsigned long timeout, +long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout, u32 __user *uaddr2, u32 val2, u32 val3) { - int ret; + int clockrt, ret = -ENOSYS; + int cmd = op & FUTEX_CMD_MASK; + int fshared = 0; + + if (!(op & FUTEX_PRIVATE_FLAG)) + fshared = 1; - switch (op) { + clockrt = op & FUTEX_CLOCK_REALTIME; + if (clockrt && cmd != FUTEX_WAIT_BITSET) + return -ENOSYS; + + switch (cmd) { case FUTEX_WAIT: - ret = futex_wait(uaddr, val, timeout); + val3 = FUTEX_BITSET_MATCH_ANY; + case FUTEX_WAIT_BITSET: + ret = futex_wait(uaddr, fshared, val, timeout, val3, clockrt); break; case FUTEX_WAKE: - ret = futex_wake(uaddr, val); - break; - case FUTEX_FD: - /* non-zero val means F_SETOWN(getpid()) & F_SETSIG(val) */ - ret = futex_fd(uaddr, val); + val3 = FUTEX_BITSET_MATCH_ANY; + case FUTEX_WAKE_BITSET: + ret = futex_wake(uaddr, fshared, val, val3); break; case FUTEX_REQUEUE: - ret = futex_requeue(uaddr, uaddr2, val, val2, NULL); + ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, NULL); break; case FUTEX_CMP_REQUEUE: - ret = futex_requeue(uaddr, uaddr2, val, val2, &val3); + ret = futex_requeue(uaddr, fshared, uaddr2, val, val2, &val3); break; case FUTEX_WAKE_OP: - ret = futex_wake_op(uaddr, uaddr2, val, val2, val3); + ret = futex_wake_op(uaddr, fshared, uaddr2, val, val2, val3); break; case FUTEX_LOCK_PI: - ret = futex_lock_pi(uaddr, val, timeout, val2, 0); + if (futex_cmpxchg_enabled) + ret = futex_lock_pi(uaddr, fshared, val, timeout, 0); break; case FUTEX_UNLOCK_PI: - ret = futex_unlock_pi(uaddr); + if (futex_cmpxchg_enabled) + ret = futex_unlock_pi(uaddr, fshared); break; case FUTEX_TRYLOCK_PI: - ret = futex_lock_pi(uaddr, 0, timeout, val2, 1); + if (futex_cmpxchg_enabled) + ret = futex_lock_pi(uaddr, fshared, 0, timeout, 1); break; default: ret = -ENOSYS; @@ -1814,59 +1980,62 @@ long do_futex(u32 __user *uaddr, int op, u32 val, unsigned long timeout, } -asmlinkage long sys_futex(u32 __user *uaddr, int op, u32 val, - struct timespec __user *utime, u32 __user *uaddr2, - u32 val3) +SYSCALL_DEFINE6(futex, u32 __user *, uaddr, int, op, u32, val, + struct timespec __user *, utime, u32 __user *, uaddr2, + u32, val3) { - struct timespec t; - unsigned long timeout = MAX_SCHEDULE_TIMEOUT; + struct timespec ts; + ktime_t t, *tp = NULL; u32 val2 = 0; + int cmd = op & FUTEX_CMD_MASK; - if (utime && (op == FUTEX_WAIT || op == FUTEX_LOCK_PI)) { - if (copy_from_user(&t, utime, sizeof(t)) != 0) + if (utime && (cmd == FUTEX_WAIT || cmd == FUTEX_LOCK_PI || + cmd == FUTEX_WAIT_BITSET)) { + if (copy_from_user(&ts, utime, sizeof(ts)) != 0) return -EFAULT; - if (!timespec_valid(&t)) + if (!timespec_valid(&ts)) return -EINVAL; - if (op == FUTEX_WAIT) - timeout = timespec_to_jiffies(&t) + 1; - else { - timeout = t.tv_sec; - val2 = t.tv_nsec; - } + + t = timespec_to_ktime(ts); + if (cmd == FUTEX_WAIT) + t = ktime_add_safe(ktime_get(), t); + tp = &t; } /* - * requeue parameter in 'utime' if op == FUTEX_REQUEUE. + * requeue parameter in 'utime' if cmd == FUTEX_REQUEUE. + * number of waiters to wake in 'utime' if cmd == FUTEX_WAKE_OP. */ - if (op == FUTEX_REQUEUE || op == FUTEX_CMP_REQUEUE) + if (cmd == FUTEX_REQUEUE || cmd == FUTEX_CMP_REQUEUE || + cmd == FUTEX_WAKE_OP) val2 = (u32) (unsigned long) utime; - return do_futex(uaddr, op, val, timeout, uaddr2, val2, val3); -} - -static int futexfs_get_sb(struct file_system_type *fs_type, - int flags, const char *dev_name, void *data, - struct vfsmount *mnt) -{ - return get_sb_pseudo(fs_type, "futex", NULL, 0xBAD1DEA, mnt); + return do_futex(uaddr, op, val, tp, uaddr2, val2, val3); } -static struct file_system_type futex_fs_type = { - .name = "futexfs", - .get_sb = futexfs_get_sb, - .kill_sb = kill_anon_super, -}; - -static int __init init(void) +static int __init futex_init(void) { - unsigned int i; + u32 curval; + int i; - register_filesystem(&futex_fs_type); - futex_mnt = kern_mount(&futex_fs_type); + /* + * This will fail and we want it. Some arch implementations do + * runtime detection of the futex_atomic_cmpxchg_inatomic() + * functionality. We want to know that before we call in any + * of the complex code paths. Also we want to prevent + * registration of robust lists in that case. NULL is + * guaranteed to fault and we get -EFAULT on functional + * implementation, the non functional ones will return + * -ENOSYS. + */ + curval = cmpxchg_futex_value_locked(NULL, 0, 0); + if (curval == -EFAULT) + futex_cmpxchg_enabled = 1; for (i = 0; i < ARRAY_SIZE(futex_queues); i++) { - INIT_LIST_HEAD(&futex_queues[i].chain); + plist_head_init(&futex_queues[i].chain, &futex_queues[i].lock); spin_lock_init(&futex_queues[i].lock); } + return 0; } -__initcall(init); +__initcall(futex_init);